SWIR Camera for Semiconductor and Wafer Inspection

The Hidden Inspection Problem: A Clean Surface Can Still Carry Risk

First, semiconductor inspection often begins with a quiet but serious question: why does a wafer that looked acceptable earlier create trouble later? In many cases, the visible image was not wrong. It was simply incomplete.

On a real inspection line, this issue may appear as sudden wafer breakage after transfer, repeated yield loss after bonding, or unstable decisions around faint scratches. Meanwhile, the visible image may still look almost normal. This gap between surface appearance and material condition is the reason short-wave infrared imaging becomes important.

For example, an internal microcrack can remain weak in visible contrast. A polished area can reflect too much light and hide a fine scratch. A wafer edge can carry small stress damage that only becomes serious after heat, contact, or movement. Therefore, the inspection station must separate real material risk from normal brightness.

However, the goal is not to add a more technical camera for its own sake. The practical goal is earlier and clearer judgment. If the imaging method reveals a hidden defect before the next process step, it can support cleaner sorting, fewer repeated checks, and more stable equipment integration.

In daily production, this difference changes the inspection mindset. A vision engineer does not only ask whether the image is bright enough. Instead, the better question is whether the defect and background separate clearly under realistic process conditions.

Therefore, the better way to discuss short-wave infrared inspection is simple: explain the phenomenon, explain the reason, give a judgment method, and then connect the inspection scene naturally to the right product path.

Why Short-Wave Infrared Sees Hidden Features

First, visible imaging mainly records the surface information that normal lighting can expose. It is effective for alignment marks, edge location, printed patterns, surface scratches, and obvious contamination. However, semiconductor materials often contain important information below that surface.

In short-wave infrared imaging, selected wavelengths interact with silicon and related materials differently. As a result, hidden cracks, weak internal structures, or material transitions may produce more useful contrast. This is why the method can support internal wafer inspection, subsurface defect review, and material-sensitive inspection steps.

In other words, short-wave infrared does not simply “see better.” It sees differently. That difference matters when the inspection target is not color, shape, or surface texture alone. Instead, the target may depend on transmission, absorption, scattering, or a subtle change inside the material.

Therefore, the practical question is not whether infrared sounds advanced. The useful question is whether the target defect creates a stable contrast difference under a suitable wavelength. If the answer is yes, the method can become a strong inspection tool. If the answer is no, another lighting method may work better.

For example, one wafer may show a hidden fracture under transmitted infrared lighting. Another sample may need reflected lighting because the defect sits near the surface. Meanwhile, a coated or bonded sample may require a different optical path because the layer structure changes how light passes through the material.

This is also why sample testing is important. A specification can show sensor capability, but it cannot prove the behavior of a specific wafer, coating, scratch, crack, or layer stack. Real samples should guide the final decision, especially when production reliability matters.

Wafer Inspection: Where Short-Wave Infrared Adds Real Value

First, wafer inspection is not one single task. It may include incoming material checks, edge review, internal crack screening, die-area inspection, bonding verification, layer-related review, and package support. Therefore, the imaging method should follow the station, not only the general word “wafer.”

In a stopped-wafer station, the inspection system has time to capture a complete frame. This can help microscope-style review, selected region analysis, and engineering sample testing. Meanwhile, a moving station needs a different approach because motion, trigger timing, and lighting stability affect every image.

Internal crack screening

For internal crack screening, transmitted short-wave infrared lighting may be tested. The reason is direct: if the material allows suitable infrared light to pass through, cracks or internal structures may change the signal. As a result, the image can show faint lines, shadows, or local gray changes that visible light may miss.

However, the result depends on the wafer, thickness, wavelength, lens, light power, and exposure. Therefore, internal crack detection should never rely only on product description. Real wafer samples should be tested before the station is finalized.

In practical terms, a useful internal crack test should include at least three sample groups. One group should include clear failures. Another group should include normal wafers. A third group should include borderline wafers that create real production uncertainty.

Then, the imaging result should be checked against the process goal. If the defect appears clearly only in one controlled lab image, the result is not enough. Instead, the system should show stable contrast across normal handling position, reasonable exposure range, and expected sample variation.

Subsurface scratch and weak surface contrast

For micro-scratch inspection, reflected lighting may be more useful. A fine scratch may scatter light differently from a smooth surface. However, the angle matters. Too much direct reflection can hide the defect, while low-angle lighting may make it stand out more clearly.

In practice, this is where many inspection trials become realistic. A sample that looks clean under one light may show a weak line under another. Therefore, the lighting angle, camera exposure, and lens should be tested as one optical system.

Also, scratch inspection should consider false positives. Polishing texture, dust, coating edge, and normal surface variation can create marks that look similar to defects. Therefore, the software rule should be tested with acceptable samples, not only failed samples.

Edge inspection and handling risk

For edge inspection, the system must control both lighting and mechanics. A small chip near the edge can create later breakage risk. Meanwhile, a curved or reflective edge can create bright glare. Therefore, fixture repeatability and background control become just as important as camera sensitivity.

Moreover, wafer edge review often depends on motion format. A stopped wafer may use area scan. A rotating wafer or moving edge station may require line scan. Therefore, the mechanical path should be discussed before product format is selected.

For edge-related stations, the most useful early question is simple: will the inspection area stay still, rotate, or move continuously? Once that answer is clear, the camera format, lighting width, trigger method, and software reconstruction method become easier to define.

Lighting: The Difference Between a Useful Image and a Pretty Image

First, lighting decides whether the camera receives useful information. A sensitive sensor cannot create reliable contrast from a poorly designed optical path. Therefore, lighting should be planned as part of the inspection method, not as an accessory added at the end.

In wafer inspection, transmitted lighting can support internal review when the material allows useful infrared passage. Meanwhile, reflected lighting can support surface scratch or contamination review. In addition, low-angle lighting can help emphasize small texture changes that normal direct lighting may flatten.

However, each lighting path carries trade-offs. Transmitted lighting may require space below the sample. Reflected lighting may create glare. Low-angle lighting may exaggerate normal texture. Therefore, the lighting decision should always come from real defect images.

In practice, a good test sequence is simple. First, image the same sample under visible light. Next, test reflected short-wave infrared illumination. Then, test transmitted illumination if the station allows it. Finally, compare whether the target defect becomes easier to separate from normal background variation.

Moreover, stable lighting makes software decisions easier. If the background changes every time the wafer moves, the algorithm must fight the setup instead of detecting the defect. Therefore, illumination uniformity, exposure stability, enclosure design, and sample positioning should be confirmed early.

Filters may also help in some systems. A filter can reduce unwanted wavelength bands and keep the useful signal cleaner. However, filter choice, light power, lens transmission, and exposure should confirm by project requirements because different materials respond differently.

Finally, lighting space should be reserved during mechanical planning. If the machine frame leaves no room for transmitted lighting or side lighting, the inspection method becomes limited before testing even begins. Therefore, optical access should be part of early station design.

Selection Check

Use the defect scene to decide whether short-wave infrared is necessary

A stable inspection plan should not begin with a model number. Instead, it should begin with the defect behavior. The following checks help decide whether infrared imaging deserves a test before the station layout is fixed.

How to Judge Whether Short-Wave Infrared Fits the Inspection Task

First, the decision should start from the defect, not the camera. If the defect is visible, stable, and easy to separate under normal light, visible inspection may remain the better solution. However, if visible imaging shows weak contrast or unstable results, short-wave infrared testing becomes worth considering.

A practical judgment method has four parts. The first part is defect type. Internal cracks, hidden scratches, edge stress, bonding voids, layer changes, and contamination do not behave the same way. Therefore, each defect type should have its own image test.

The second part is material behavior. Silicon, coating, adhesive, ceramic, package material, and surface finish can all respond differently. Therefore, sample material should be described clearly before the test begins.

The third part is station motion. A stopped wafer can allow longer exposure and full-frame capture. Meanwhile, a moving line needs tighter control of exposure, trigger, scan timing, and light intensity. Therefore, motion format can change the camera type.

The fourth part is software stability. A defect must not only appear in one good image. It must remain separable across normal wafers, borderline defects, surface variation, lighting drift, and production speed. Therefore, the optical test and algorithm test should happen together.

Next, the pass and fail rule should be written before the final model is selected. If the station only needs to flag suspected cracks for review, one level of image quality may be enough. If the station must make direct pass or fail decisions, the contrast and repeatability requirement becomes stricter.

Also, the integration team should check what happens after detection. Some systems need only image capture. Others need trigger output, result logging, image storage, PLC communication, or connection with existing machine vision software. Therefore, software workflow should be part of the selection process.

In short, the correct question is not “Which model has the strongest parameter?” The better question is “Which imaging path creates stable contrast for the target defect under real station conditions?” This question leads to a more reliable selection process.

Product Fit: From Inspection Scene to MindVision Product Path

First, the main product entry for this topic should remain the short-wave infrared product category. The page supports semiconductor-oriented discussion because it focuses on short-wave infrared imaging, microscopic imaging, and integration with machine vision workflows. Therefore, the article should naturally guide the reader to the short-wave infrared camera category when the content moves from explanation to product review.

Meanwhile, the broader industrial camera product range should support wider selection. Some semiconductor lines need more than short-wave infrared imaging. They may also need visible area scan cameras, line scan cameras, smart camera integration, lighting, lenses, or customized support.

For company background and product support, the anchor MindVision industrial camera manufacturer can be used once in a natural place. This keeps the article focused while still passing authority to the homepage.

Area-scan path: fixed wafer, microscope view, selected region

An area-scan short-wave infrared product path is suitable when the sample stops during capture. This makes it easier to review internal cracks, selected die areas, microscope images, or material contrast at a fixed position. Moreover, it allows slower setup work during feasibility testing.

However, area scan is not automatically better. The field of view, defect size, lens magnification, exposure time, frame rate, and mounting space should confirm by project requirements. If a large field needs fine detail, optics and resolution must be checked together. If motion blur appears, exposure and lighting power must be reviewed.

Line-scan path: moving edge, strip, or continuous inspection

A line-scan product path becomes more relevant when the material moves continuously. The camera records one line at a time while motion builds the full image. Therefore, this method can support wafer edge scanning, strip-like material inspection, or conveyor-based review.

However, line scan depends heavily on motion control. The encoder, stage speed, illumination width, scan direction, and trigger logic must work together. If motion is unstable, the image can stretch, compress, or lose uniformity. Therefore, line scan should be treated as a full system decision.

Continuous Inspection

Line-scan product path for moving wafer edges or long surfaces

When the inspection object moves, a line-scan format can build a continuous image across the moving direction. This is useful for wafer edge scanning, strip-like material inspection, and long-area review.

Final use should confirm by project requirements, especially scan width, motion speed, light uniformity, encoder timing, and software reconstruction.

View Line-scan Product

In addition, the short-wave infrared imaging introduction page can support readers who need concept-level background. The application article on internal wafer inspection with short-wave infrared imaging is more suitable when the focus is hidden wafer defect detection.

Therefore, the internal linking path should be clear. The article explains the inspection problem, the introduction page supports concept learning, the case article supports application proof, and the product landing page guides technical selection.

Comparison Table: Match the Method to the Wafer Inspection Scene

Instead of comparing products only by specification, the following table compares inspection scenes. This is closer to real engineering decisions because every station has its own defect target, motion format, and lighting limits.

This table should be treated as a planning tool. Real samples may behave differently from expectation. Therefore, the final choice should always come from image trials, algorithm review, and integration constraints.

Also, the table helps avoid a common misunderstanding. The best method is not always the most advanced one. Instead, the best method is the one that gives stable contrast, fits the machine layout, and supports reliable software decisions.

Common Mistakes That Make Infrared Inspection Look Unstable

First, some projects start with the highest resolution they can find. However, resolution does not create defect contrast by itself. If the defect remains weak under the chosen lighting, a larger image simply becomes a larger uncertain image.

Second, lighting is sometimes tested too late. This creates mechanical problems because the final fixture may leave no room for transmitted light, side light, or filters. Therefore, lighting space should be reserved before the station layout becomes fixed.

Third, some inspection plans use only ideal samples. This makes the early result look clean, but production variation may expose weakness later. Therefore, borderline samples and normal variation should appear in the first test set.

Fourth, the software threshold is sometimes chosen from one image. However, semiconductor inspection needs repeatability. A useful threshold should survive normal brightness changes, wafer differences, and fixture tolerance.

Fifth, product choice is sometimes separated from lens choice. This creates avoidable problems because the lens must support the required wavelength, working distance, field of view, and image sharpness. Therefore, lens and camera should be evaluated together.

Finally, system integration can be underestimated. Camera interface, cable routing, trigger timing, lens mount, working distance, enclosure reflection, and software compatibility all affect stability. Therefore, selection should remain a system decision.

Extended Reading

For deeper reading, the following internal pages support the same topic without distracting from the main semiconductor and wafer inspection focus.

FAQ

Conclusion: Start With the Defect, Then Select the Imaging Path

In semiconductor and wafer inspection, the most reliable decision starts from the defect scene. A clean visible image does not always mean the material is safe. Therefore, short-wave infrared inspection should be considered when hidden cracks, weak scratches, edge stress, or layer-related changes require contrast beyond visible light.

The practical path is clear. First, define the defect. Next, test the lighting. Then, compare real sample images. Finally, match the product format to the station. For broader industrial camera support and semiconductor vision discussion, a SWIR camera project can start from MindVision’s homepage before moving into the short-wave infrared product category.

• Prepare real pass, fail, and borderline wafer samples before selection.
• Test transmitted, reflected, and angled lighting before fixing the station layout.
• Confirm model, lens, wavelength, exposure, trigger, and software by project requirements.